DOCSLIB.ORG

UNIT 10C Multitasking & Operating Systems

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
UNIT 10C Multitasking & Operating Systems 11/5/2011 UNIT 10C Concurrency: Multitasking & Deadlock 15110 Principles of Computing, Carnegie 1 Mellon University - CORTINA Multitasking & Operating Systems • Multitasking - The coordination of several computational processes on one processor. • An operating system (OS) is the system software responsible for the direct control and management of hardware and basic system operations. • An OS provides a foundation upon which to run application software such as word processing programs, web browsers, etc. from Wikipedia 15110 Principles of Computing, Carnegie 2 Mellon University - CORTINA 1 11/5/2011 Operating System “Flavors” • UNIX – System V, BSD, Linux – Proprietary: Solaris, Mac OS X • Windows – Windows XP – Windows Vista – Windows 7 15110 Principles of Computing, Carnegie 3 Mellon University - CORTINA Multitasking • Each (un-blocked) application runs for a very short time on the computer's processor and then another process runs, then another... • This is done at such a fast rate that it appears to the computer user that all applications are running at the same time. – How do actors appear to move in a motion picture? 15110 Principles of Computing, Carnegie 4 Mellon University - CORTINA 2 11/5/2011 Critical Section • A critical section is a section of computer code that must only be executed by one process or thread at a time. • Examples: – A printer queue receiving a file to be printed – Code to set a seat as reserved – Web server that generates and returns a web page showing current registration in course 15110 Principles of Computing, Carnegie 5 Mellon University - CORTINA A Critical Section Cars may not make turns through this intersection. They may only proceed straight ahead. When a car stops at this intersection, it can proceed straight ahead if there is no car to its left and no car to its right. Otherwise it must wait some random amount of seconds and then check again to see if it's safe to proceed. If not, it waits again, and so on. 15110 Principles of Computing, Carnegie 6 Mellon University - CORTINA 3 11/5/2011 Shared Computing Resources • memory • tape drives • disk drives • printers • communication ports • input devices (keyboard, mouse) 15110 Principles of Computing, Carnegie 7 Mellon University - CORTINA Deadlock • Deadlock is the condition when two or more processes are all waiting for some shared resource that other processes of the group hold, causing all processes to wait forever without proceeding. • How can deadlock occur at the intersection with the 4-way stop? 15110 Principles of Computing, Carnegie 8 Mellon University - CORTINA 4 11/5/2011 Dining Philosopher’s Problem Socrates Aristotle Plato Homer 15110 Principles of Computing, Carnegie 9 Mellon University - CORTINA The Dining Philosopher’s • Each philosopher thinks for a while, then picks up his left fork, then picks up his right fork, then eats, then puts down his left fork, then puts down his right fork, thinks for a while... – We assume here that each philosopher thinks and eats for random times, and a philosopher cannot be interrupted while he picks up or puts down a single fork. • Each fork models a "resource" on a computer controlled by an OS. • Original problem proposed by Edsgar Dijkstra. 15110 Principles of Computing, Carnegie 10 Mellon University - CORTINA 5 11/5/2011 Dining Philosopher’s Problem • There are N philosophers. • Philosopher i does the following: 1. THINK N=4 2. Pick up fork i. Fork 3 Fork 2 3. Pick up fork (i+1) modulo N. 2 4. EAT 3 1 5. Put down fork i. 0 6. Put down fork (i+1) modulo N. Fork 0 Fork 1 7. Go to step 1. NOTE: (i+1) modulo N = i+1 , if 0 < i < N-1 (i+1) modulo N = 0, if i = N-1 15110 Principles of Computing, Carnegie 11 Mellon University - CORTINA Dining Philosopher’s Problem • There are N philosophers. • Philosopher i does the following: 1. THINK N=4 2. Pick up fork i. Fork 3 Fork 2 3. Pick up fork (i+1) modulo N. 2 4. EAT 3 1 5. Put down fork i. 0 6. Put down fork (i+1) modulo N. Fork 0 Fork 1 7. Go to step 1. How can deadlock occur here? 15110 Principles of Computing, Carnegie 12 Mellon University - CORTINA 6 11/5/2011 Removing Deadlock • Philosopher i performs the following: 1. THINK 2. If i is not equal to N-1: a. Pick up fork i b . Pick up fork i+1 N=4 3. If i equals N-1: a. Pick up fork 0 Fork 3 Fork 2 b . Pick up fork N-1 2 4. EAT 5. If i is not equal to N-1: 3 1 a. Put down fork i 0 b . Put down fork i+1 Fork 0 Fork 1 6. If i equals N-1: a. Put down fork 0 b . Put down fork N-1 Can a philosopher still starve? 7. Goto step 1 15110 Principles of Computing, Carnegie 13 Mellon University - CORTINA Semaphores • A (binary) semaphore S is a shared variable that holds an integer that is initialized to 1. • We can use a semaphore to protect a critical section so that only one process accesses this section at a given time: Initially, semaphore S is 1. The non-critical program code first process to request S sets S to 0, blocking other processes Request S from proceeding once they get to their request operations. Once CRITICAL SECTION the first process finishes running the critical section, it releases S, Release S causing S to increase to 1, non-critical program code allowing another process to proceed into the critical section. 15110 Principles of Computing, Carnegie 14 Mellon University - CORTINA 7 11/5/2011 Summary • Concurrent computing and multitasking are very challenging problems in computer science. • The simple rules of sequential computation no longer apply when we compute – in a distributed manner across networks – in a single computer system in a multitasking environment 15110 Principles of Computing, Carnegie 15 Mellon University - CORTINA 8.
Load more

Recommended publications
  • Parallel Programming
    Parallel Programming Parallel Programming Parallel Computing Hardware Shared memory: multiple cpus are attached to the BUS all processors share the same primary memory the same memory address on different CPU’s refer to the same memory location CPU-to-memory connection becomes a bottleneck: shared memory computers cannot scale very well Parallel Programming Parallel Computing Hardware Distributed memory: each processor has its own private memory computational tasks can only operate on local data infinite available memory through adding nodes requires more difficult programming Parallel Programming OpenMP versus MPI OpenMP (Open Multi-Processing): easy to use; loop-level parallelism non-loop-level parallelism is more difficult limited to shared memory computers cannot handle very large problems MPI(Message Passing Interface): require low-level programming; more difficult programming scalable cost/size can handle very large problems Parallel Programming MPI Distributed memory: Each processor can access only the instructions/data stored in its own memory. The machine has an interconnection network that supports passing messages between processors. A user specifies a number of concurrent processes when program begins. Every process executes the same program, though theflow of execution may depend on the processors unique ID number (e.g. “if (my id == 0) then ”). ··· Each process performs computations on its local variables, then communicates with other processes (repeat), to eventually achieve the computed result. In this model, processors pass messages both to send/receive information, and to synchronize with one another. Parallel Programming Introduction to MPI Communicators and Groups: MPI uses objects called communicators and groups to define which collection of processes may communicate with each other.
    [Show full text]
  • Real-Time Performance During CUDA™ a Demonstration and Analysis of Redhawk™ CUDA RT Optimizations
    A Concurrent Real-Time White Paper 2881 Gateway Drive Pompano Beach, FL 33069 (954) 974-1700 www.concurrent-rt.com Real-Time Performance During CUDA™ A Demonstration and Analysis of RedHawk™ CUDA RT Optimizations By: Concurrent Real-Time Linux® Development Team November 2010 Overview There are many challenges to creating a real-time Linux distribution that provides guaranteed low process-dispatch latencies and minimal process run-time jitter. Concurrent Real Time’s RedHawk Linux distribution meets and exceeds these challenges, providing a hard real-time environment on many qualified hardware configurations, even in the presence of a heavy system load. However, there are additional challenges faced when guaranteeing real-time performance of processes while CUDA applications are simultaneously running on the system. The proprietary CUDA driver supplied by NVIDIA® frequently makes demands upon kernel resources that can dramatically impact real-time performance. This paper discusses a demonstration application developed by Concurrent to illustrate that RedHawk Linux kernel optimizations allow hard real-time performance guarantees to be preserved even while demanding CUDA applications are running. The test results will show how RedHawk performance compares to CentOS performance running the same application. The design and implementation details of the demonstration application are also discussed in this paper. Demonstration This demonstration features two selectable real-time test modes: 1. Jitter Mode: measure and graph the run-time jitter of a real-time process 2. PDL Mode: measure and graph the process-dispatch latency of a real-time process While the demonstration is running, it is possible to switch between these different modes at any time.
    [Show full text]
  • Unit: 4 Processes and Threads in Distributed Systems
    Unit: 4 Processes and Threads in Distributed Systems Thread A program has one or more locus of execution. Each execution is called a thread of execution. In traditional operating systems, each process has an address space and a single thread of execution. It is the smallest unit of processing that can be scheduled by an operating system. A thread is a single sequence stream within in a process. Because threads have some of the properties of processes, they are sometimes called lightweight processes. In a process, threads allow multiple executions of streams. Thread Structure Process is used to group resources together and threads are the entities scheduled for execution on the CPU. The thread has a program counter that keeps track of which instruction to execute next. It has registers, which holds its current working variables. It has a stack, which contains the execution history, with one frame for each procedure called but not yet returned from. Although a thread must execute in some process, the thread and its process are different concepts and can be treated separately. What threads add to the process model is to allow multiple executions to take place in the same process environment, to a large degree independent of one another. Having multiple threads running in parallel in one process is similar to having multiple processes running in parallel in one computer. Figure: (a) Three processes each with one thread. (b) One process with three threads. In former case, the threads share an address space, open files, and other resources. In the latter case, process share physical memory, disks, printers and other resources.
    [Show full text]
  • Gpu Concurrency
    GPU CONCURRENCY ROBERT SEARLES 5/26/2021 EXECUTION SCHEDULING & MANAGEMENT Pre-emptive scheduling Concurrent scheduling Processes share GPU through time-slicing Processes run on GPU simultaneously Scheduling managed by system User creates & manages scheduling streams C B A B C A B A time time time- slice 2 CUDA CONCURRENCY MECHANISMS Streams MPS MIG Partition Type Single process Logical Physical Max Partitions Unlimited 48 7 Performance Isolation No By percentage Yes Memory Protection No Yes Yes Memory Bandwidth QoS No No Yes Error Isolation No No Yes Cross-Partition Interop Always IPC Limited IPC Reconfigure Dynamic Process launch When idle MPS: Multi-Process Service MIG: Multi-Instance GPU 3 CUDA STREAMS 4 STREAM SEMANTICS 1. Two operations issued into the same stream will execute in issue- order. Operation B issued after Operation A will not begin to execute until Operation A has completed. 2. Two operations issued into separate streams have no ordering prescribed by CUDA. Operation A issued into stream 1 may execute before, during, or after Operation B issued into stream 2. Operation: Usually, cudaMemcpyAsync or a kernel call. More generally, most CUDA API calls that take a stream parameter, as well as stream callbacks. 5 STREAM EXAMPLES Host/Device execution concurrency: Kernel<<<b, t>>>(…); // this kernel execution can overlap with cpuFunction(…); // this host code Concurrent kernels: Kernel<<<b, t, 0, streamA>>>(…); // these kernels have the possibility Kernel<<<b, t, 0, streamB>>>(…); // to execute concurrently In practice, concurrent
    [Show full text]
  • The Challenges of Hardware Synthesis from C-Like Languages
    The Challenges of Hardware Synthesis from C-like Languages Stephen A. Edwards∗ Department of Computer Science Columbia University, New York Abstract most successful C-like languages, in fact, bear little syntactic or semantic resemblance to C, effectively forcing users to learn The relentless increase in the complexity of integrated circuits a new language anyway. As a result, techniques for synthesiz- we can fabricate imposes a continuing need for ways to de- ing hardware from C either generate inefficient hardware or scribe complex hardware succinctly. Because of their ubiquity propose a language that merely adopts part of C syntax. and flexibility, many have proposed to use the C and C++ lan- For space reasons, this paper is focused purely on the use of guages as specification languages for digital hardware. Yet, C-like languages for synthesis. I deliberately omit discussion tools based on this idea have seen little commercial interest. of other important uses of a design language, such as validation In this paper, I argue that C/C++ is a poor choice for specify- and algorithm exploration. C-like languages are much more ing hardware for synthesis and suggest a set of criteria that the compelling for these tasks, and one in particular (SystemC) is next successful hardware description language should have. now widely used, as are many ad hoc variants. 1 Introduction 2 A Short History of C Familiarity is the main reason C-like languages have been pro- Dennis Ritchie developed C in the early 1970 [18] based on posed for hardware synthesis. Synthesize hardware from C, experience with Ken Thompson’s B language, which had itself proponents claim, and we will effectively turn every C pro- evolved from Martin Richards’ BCPL [17].
    [Show full text]
  • Hyper-Threading Technology Architecture and Microarchitecture
    Hyper-Threading Technology Architecture and Microarchitecture Deborah T. Marr, Desktop Products Group, Intel Corp. Frank Binns, Desktop ProductsGroup, Intel Corp. David L. Hill, Desktop Products Group, Intel Corp. Glenn Hinton, Desktop Products Group, Intel Corp. David A. Koufaty, Desktop Products Group, Intel Corp. J. Alan Miller, Desktop Products Group, Intel Corp. Michael Upton, CPU Architecture, Desktop Products Group, Intel Corp. Index words: architecture, microarchitecture, Hyper-Threading Technology, simultaneous multi- threading, multiprocessor INTRODUCTION ABSTRACT The amazing growth of the Internet and telecommunications is powered by ever-faster systems Intel’s Hyper-Threading Technology brings the concept demanding increasingly higher levels of processor of simultaneous multi-threading to the Intel performance. To keep up with this demand we cannot Architecture. Hyper-Threading Technology makes a rely entirely on traditional approaches to processor single physical processor appear as two logical design. Microarchitecture techniques used to achieve processors; the physical execution resources are shared past processor performance improvement–super- and the architecture state is duplicated for the two pipelining, branch prediction, super-scalar execution, logical processors. From a software or architecture out-of-order execution, caches–have made perspective, this means operating systems and user microprocessors increasingly more complex, have more programs can schedule processes or threads to logical transistors, and consume more power. In fact, transistor processors as they would on multiple physical counts and power are increasing at rates greater than processors. From a microarchitecture perspective, this processor performance. Processor architects are means that instructions from both logical processors therefore looking for ways to improve performance at a will persist and execute simultaneously on shared greater rate than transistor counts and power execution resources.
    [Show full text]
  • Scheduling Many-Task Workloads on Supercomputers: Dealing with Trailing Tasks
    Scheduling Many-Task Workloads on Supercomputers: Dealing with Trailing Tasks Timothy G. Armstrong, Zhao Zhang Daniel S. Katz, Michael Wilde, Ian T. Foster Department of Computer Science Computation Institute University of Chicago University of Chicago & Argonne National Laboratory [email protected], [email protected] [email protected], [email protected], [email protected] Abstract—In order for many-task applications to be attrac- as a worker and allocate one task per node. If tasks are tive candidates for running on high-end supercomputers, they single-threaded, each core or virtual thread can be treated must be able to benefit from the additional compute, I/O, as a worker. and communication performance provided by high-end HPC hardware relative to clusters, grids, or clouds. Typically this The second feature of many-task applications is an empha- means that the application should use the HPC resource in sis on high performance. The many tasks that make up the such a way that it can reduce time to solution beyond what application effectively collaborate to produce some result, is possible otherwise. Furthermore, it is necessary to make and in many cases it is important to get the results quickly. efficient use of the computational resources, achieving high This feature motivates the development of techniques to levels of utilization. Satisfying these twin goals is not trivial, because while the efficiently run many-task applications on HPC hardware. It parallelism in many task computations can vary over time, allows people to design and develop performance-critical on many large machines the allocation policy requires that applications in a many-task style and enables the scaling worker CPUs be provisioned and also relinquished in large up of existing many-task applications to run on much larger blocks rather than individually.
    [Show full text]
  • Intel's Hyper-Threading
    Intel’s Hyper-Threading Cody Tinker & Christopher Valerino Introduction ● How to handle a thread, or the smallest portion of code that can be run independently, plays a core component in enhancing a program's parallelism. ● Due to the fact that modern computers process many tasks and programs simultaneously, techniques that allow for threads to be handled more efficiently to lower processor downtime are valuable. ● Motive is to reduce the number of idle resources of a processor. ● Examples of processors that use hyperthreading ○ Intel Xeon D-1529 ○ Intel i7-6785R ○ Intel Pentium D1517 Introduction ● Two main trends have been followed to increase parallelism: ○ Increase number of cores on a chip ○ Increase core throughput What is Multithreading? ● Executing more than one thread at a time ● Normally, an operating system handles a program by scheduling individual threads and then passing them to the processor. ● Two different types of hardware multithreading ○ Temporal ■ One thread per pipeline stage ○ Simultaneous (SMT) ■ Multiple threads per pipeline stage ● Intel’s hyper-threading is a SMT design Hyper-threading ● Hyper-threading is the hardware solution to increasing processor throughput by decreasing resource idle time. ● Allows multiple concurrent threads to be executed ○ Threads are interleaved so that resources not being used by one thread are used by others ○ Most processors are limited to 2 concurrent threads per physical core ○ Some do support 8 concurrent threads per physical core ● Needs the ability to fetch instructions from
    [Show full text]
  • COSC 6385 Computer Architecture - Multi-Processors (IV) Simultaneous Multi-Threading and Multi-Core Processors Edgar Gabriel Spring 2011
    COSC 6385 Computer Architecture - Multi-Processors (IV) Simultaneous multi-threading and multi-core processors Edgar Gabriel Spring 2011 Edgar Gabriel Moore’s Law • Long-term trend on the number of transistor per integrated circuit • Number of transistors double every ~18 month Source: http://en.wikipedia.org/wki/Images:Moores_law.svg COSC 6385 – Computer Architecture Edgar Gabriel 1 What do we do with that many transistors? • Optimizing the execution of a single instruction stream through – Pipelining • Overlap the execution of multiple instructions • Example: all RISC architectures; Intel x86 underneath the hood – Out-of-order execution: • Allow instructions to overtake each other in accordance with code dependencies (RAW, WAW, WAR) • Example: all commercial processors (Intel, AMD, IBM, SUN) – Branch prediction and speculative execution: • Reduce the number of stall cycles due to unresolved branches • Example: (nearly) all commercial processors COSC 6385 – Computer Architecture Edgar Gabriel What do we do with that many transistors? (II) – Multi-issue processors: • Allow multiple instructions to start execution per clock cycle • Superscalar (Intel x86, AMD, …) vs. VLIW architectures – VLIW/EPIC architectures: • Allow compilers to indicate independent instructions per issue packet • Example: Intel Itanium series – Vector units: • Allow for the efficient expression and execution of vector operations • Example: SSE, SSE2, SSE3, SSE4 instructions COSC 6385 – Computer Architecture Edgar Gabriel 2 Limitations of optimizing a single instruction
    [Show full text]
  • Synchronously Waiting on Asynchronous Operations
    Document No. P1171R0 Date 2018-10-07 Reply To Lewis Baker <[email protected]> Audience SG1, LEWG Synchronously waiting on asynchronous operations Overview The paper P1056R0 introduces a new std::experimental::task<T> type. This type represents an ​ ​ ​ ​ ​ asynchronous operation that requires applying operator co_await to the task retrieve the result. The task<T> ​ type is an instance of the more general concept of Awaitable types. The limitation of Awaitable types is that they can only be co_awaited from within a coroutine. For example, from the body of another task<T> coroutine. However, then you still end up with another Awaitable type that must be co_awaited within another coroutine. This presents a problem of how to start executing the first task and wait for it to complete. task<int> f(); task<int> g() { // Call to f() returns a not-yet-started task. // Applying operator co_await() to the task starts its execution // and suspends the current coroutine until it completes. int a = co_await f(); co_return a + 1; } int main() { task<int> t = g(); // But how do we start executing g() and waiting for it to complete // when outside of a coroutine context, such as in main()? int x = ???; return x; } This paper proposes a new function, sync_wait(), that will allow a caller to pass an arbitrary Awaitable type ​ ​ into the function. The function will co_await the passed Awaitable object on the current thread and then block ​ ​ waiting for that co_await operation to complete; either synchronously on the current thread, or ​ ​ asynchronously on another thread. When the operation completes, the result is captured on whichever thread the operation completed on.
    [Show full text]
  • Lock-Free Programming
    Lock-Free Programming Geoff Langdale L31_Lockfree 1 Desynchronization ● This is an interesting topic ● This will (may?) become even more relevant with near ubiquitous multi-processing ● Still: please don’t rewrite any Project 3s! L31_Lockfree 2 Synchronization ● We received notification via the web form that one group has passed the P3/P4 test suite. Congratulations! ● We will be releasing a version of the fork-wait bomb which doesn't make as many assumptions about task id's. – Please look for it today and let us know right away if it causes any trouble for you. ● Personal and group disk quotas have been grown in order to reduce the number of people running out over the weekend – if you try hard enough you'll still be able to do it. L31_Lockfree 3 Outline ● Problems with locking ● Definition of Lock-free programming ● Examples of Lock-free programming ● Linux OS uses of Lock-free data structures ● Miscellanea (higher-level constructs, ‘wait-freedom’) ● Conclusion L31_Lockfree 4 Problems with Locking ● This list is more or less contentious, not equally relevant to all locking situations: – Deadlock – Priority Inversion – Convoying – “Async-signal-safety” – Kill-tolerant availability – Pre-emption tolerance – Overall performance L31_Lockfree 5 Problems with Locking 2 ● Deadlock – Processes that cannot proceed because they are waiting for resources that are held by processes that are waiting for… ● Priority inversion – Low-priority processes hold a lock required by a higher- priority process – Priority inheritance a possible solution L31_Lockfree
    [Show full text]
  • IMTEC-91-58 High-Performance Computing: Industry Uses Of
    ‘F ---l_l,__*“.^*___l.ll_l-_.. - -.”_-._ .----._..-_.. ___..-.-..,_.. -...-I_.__...... liitiltd l_--._-_.---..-_---. Sl;tl,~s (;~~ttc~ral Accottttt.irtg Offiw ---- GAO 1Zcport 1,oCongressional Request,ers i HIGH-PERFORMANCE COMPUTING Industry Uses of Supercomputers and High-Speed Networks 144778 RELEASED (;AO/IM’l’E(:-!tl-T,El .-. I .II ._. ._._. “..__-I ._.,__ _. _ .._.. ._.-._ . .I .._..-.-...-..--.-- ~-- United States General Accounting Office GAO Washington, D.C. 20648 Information Management and Technology Division B-244488 July 30,199l The Honorable Ernest F. Hollings ‘Chairman, Senate Committee on Commerce, Science, and Transportation The Honorable Al Gore Chairman, Subcommittee on Science, Technology, and Space Senate Committee on Commerce, Science, and Transportation The Honorable George E. Brown, Jr. Chairman, House Committee on Science, Space, and Technology The Honorable Robert S. Walker Ranking Minority Member House Committee on Science, Space, and Technology The HonorableTim Valentine Chairman, Subcommittee on Technology and Competitiveness House Committee on Scientie, Space, and Technology The Honorable Tom Lewis Ranking Minority Member Subcommittee on Technology and Competitiveness House Committee on Science, Space, and Technology This report responds to your October 2,1990, and March 11,1991, requests for information on supercomputers and high-speed networks. You specifically asked that we l provide examples of how various industries are using supercomputers to improve products, reduce costs, save time, and provide other benefits; . identify barriers preventing the increased use of supercomputers; and . provide examples of how certain industries are using and benefitting from high-speed networks. Page 1 GAO/JMTEG91-59 Supercomputera and High-Speed Networks B244488 As agreed with the Senate Committee on Commerce, Science, and Trans- portation, and Subcommittee on Science, Technology, and Space, our review of supercomputers examined five industries-oil, aerospace, automobile, and chemical and pharmaceutical.
    [Show full text]